This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 46.30.132.228

This content was downloaded on 12/06/2014 at 11:13

Please note that terms and conditions apply.

First comprehensive inter-comparison of aerosol electrometers for particle sizes up to 200 nm

and concentration range 1000 cm−3 to 17 000 cm−3

View the table of contents for this issue, or go to the journal homepage for more

2014 Metrologia 51 293

(http://iopscience.iop.org/0026-1394/51/3/293)

Home Search Collections Journals About Contact us My IOPscience

| Bureau International des Poids et Mesures Metrologia

Metrologia 51 (2014) 293–303 doi:10.1088/0026-1394/51/3/293

First comprehensive inter-comparison ofaerosol electrometers for particle sizes upto 200 nm and concentration range1000 cm−3 to 17 000 cm−3

Richard Hogstrom1,10, Paul Quincey2, Dimitris Sarantaridis2, Felix Luond3,Andreas Nowak4, Francesco Riccobono5, Thomas Tuch6, Hiromu Sakurai7,Miles Owen8, Martti Heinonen1, Jorma Keskinen9 and Jaakko Yli-Ojanpera9,10

1 Centre for Metrology and Accreditation (MIKES), Tekniikantie 1, FI-02151 Espoo, Finland2 National Physical Laboratory (NPL), Hampton Road, Teddington TW11 0LW, UK3 Federal Institute of Metrology (METAS), Lindenweg 50, CH-3003 Bern-Wabern, Switzerland4 Physikalisch-Technische Bundesanstalt (PTB), Bundesallee 100, D-38116 Braunschweig, Germany5 European Commission, Joint Research Centre (JRC), Via E Fermi 2749, I-21027 Ispra, Italy6 Leibniz Institute fur Tropospharenforschung (TROPOS), Permoserstraße 15, D-04318 Leipzig, Germany7 National Metrology Institute of Japan, National Institute of Advanced Industrial Science and Technology (AIST),1-1-1 Umezono, Tsukuba, Ibaraki, Japan8 US Army Primary Standards Laboratory (APSL), Bldg 5435 Fowler Rd, Redstone Arsenal, AL 35898, USA9 Tampere University of Technology (TUT), Department of Physics, Aerosol Physics Laboratory,Korkeakoulunkatu 3, FI-33101 Tampere, Finland

E-mail: richard.hogstrom@mikes.fi and jaakko.yli-ojanpera@tut.fi

Received 10 December 2013, revised 21 March 2014Accepted for publication 21 March 2014Published 28 May 2014

AbstractThe concentration of nanometre-sized particles is frequently measured in terms of particle numberconcentration using well-established measuring instruments, e.g. condensation particle counters.Traceability for these measurements can be achieved by means of calibrations using an aerosolelectrometer (AE) as a reference. A number of national metrology institutes (NMIs) and expertlaboratories provide such calibrations, but the metrological basis is at present not well establishedbecause the equivalence between the unit realizations has not been investigated by means of aninter-laboratory comparison. This paper presents the results of the first comprehensive comparison ofAEs involving NMIs and expert laboratories worldwide. The comparison covered the particle size andcharge concentrations ranges 20 nm to 200 nm and 0.16 × 10−15 C cm−3 to 2.72 × 10−15 C cm−3

(equivalent to 1000 cm−3 to 17 000 cm−3 singly charged particles), respectively. The obtained resultsagreed to within about ±3%, which was within stated uncertainties, with only a few exceptions, such asat low concentrations. Additional measurements with sub-20 nm particles show that comparisons in thissize range are more challenging and require special considerations, though agreement to within about±5% was still found with 6 nm particles. This comparison is the first and vital step towardsinternationally recognized SI traceability in particle number concentration measurements.

Keywords: comparison, particle number concentration, aerosols, electrometer

1. Introduction

Aerosol particles play an important role in many areas oftechnology and science, e.g. global climate change [1], health

10 Authors to whom any correspondence should be addressed.

effects of particulate emissions [2] and nanotechnology [3]. Inaddition, aerosol particle measurements are widely applied,e.g. in clean room and process technology, ambient airmonitoring, as well as in studying emissions from traffic,industry and power production. In recent years the interestin fine and ultrafine particles has increased, as it has been

0026-1394/14/030293+11$33.00 293 © 2014 BIPM & IOP Publishing Ltd Printed in the UK

Metrologia 51 (2014) 293 R Hogstrom et al

shown that these nanometre-sized particles have adverse healtheffects [4] and are responsible for particle formation in theatmosphere [5]. The behaviour and effect of nano-sizedparticles is better described in terms of number concentration,surface area and size distribution rather than simply massconcentration. Moreover, the sensitivity of particle numberand surface area concentration measurements is superior tomass concentration measurements owing to the negligiblemass of nano-sized particles. Measurement of particle numberis, however, preferred over surface area in the regulatorycontext as it is more clearly defined than surface area, whichis ambiguous for aerosols with complex surface structure.

Instruments measuring particle number concentration,e.g. condensation particle counters (CPCs [6]) and opticalparticle counters (OPCs [7]), are well developed and havebeen available on the market for a few decades. Numberconcentration measurements are also often integral to particlesize distribution measurements, such as when using a scanningmobility particle sizer (SMPS [8]) or an electrical lowpressure impactor (ELPI [9]). Aerosol particle numberconcentration measurements are frequently applied in cleanroom monitoring where the measurements are covered bystandards [10]. Moreover, particle number concentration hasrecently featured in the European vehicle emission legislation[11] and it is becoming increasingly important in otherareas such as ambient air monitoring. This has generated agrowing demand for traceable particle number concentrationmeasurements. Currently some national metrology institutes(NMIs) and research laboratories have primary particle numberconcentration standards [12–14]. However, the metrologicalbasis is at present incomplete because the equivalence of theactual unit realizations has not been investigated by means ofinter-laboratory comparisons.

Most of the standards in use comprise an aerosolparticle generator and an aerosol electrometer (AE). Theparticle generator provides calibration aerosols with knownsize and average charge, whereas the AE measures chargeconcentration of the generated aerosol particles. Thetraceability for charge concentration is achieved by SI traceablemeasurement of electric current induced by charged particlesinside a Faraday cup (FCUP) and the volume flow of theaerosol. If the charge of the particles from the generatoris known, the number concentration (expressed typicallyin cm−3) is directly derived from the charge concentration(e.g. in C cm−3). As a primary standard, however, theperformance of an AE contributes to the uncertainty of allparticle concentration measurements traceable through it.Therefore, a thorough validation comprising a full uncertaintyanalysis and inter-laboratory comparisons is essential.

In an AE, random noise in the current measurementbecomes the prominent source of uncertainty when measuringlow particle concentrations (low currents), thus setting a lowerlimit for the measurement range. In addition, the averagecharge of the particles must be known and contributes tothe uncertainty of the number concentration. Despite theselimitations the AE is considered the prime choice for realizingSI traceability for aerosol particle number concentration. TheAE is well suited for calibrating CPCs (an instrument type

widely in use for measuring particle number concentration inthe size range from a few nanometres to a few micrometres)because its response, after correcting for the diffusion lossesat the inlet, remains constant at about unity for particles in thesize range from 1 nm to 30 nm where the detection efficiencyof CPCs drops steeply. In fact, the draft ISO standard ISO/DIS27891 [15] describes a calibration procedure for CPCs byreference to an AE, and explicitly describes the role of NMIsin providing certification for AEs.

Up to now, only one inter-laboratory comparison onparticle number concentration and particle size distributionmeasurements has been carried out at the NMI level [16].It was performed using CPCs and SMPS which have beenindependently calibrated by the participating laboratories.This comparison was performed using combustion aerosolparticles and covered the particle size range from 50 nmto 170 nm. However, the comparison involved secondarystandards and it did not cover the sub-50 nm particle size rangewhich is relevant for many particle number measurementsusing CPCs. Recently, a bilateral comparison between theFinnish and Japanese particle number concentration standards[12, 13] was performed at the National Institute of AdvancedIndustrial Science and Technology (Japan) by calibrating asingle CPC in a wide particle size range from 10 nm to10 µm using three different calibration standards [17]. Thiscomparison was exceptional because the Finnish standardwas transported to Japan where the comparison took placeallowing direct comparison without transporting the CPC.This approach, however, is not applicable in general sincemost particle number concentration standards are not suitablefor transportation. To overcome this issue but still keepingthe influence of the comparison method at minimum, itwas agreed in the European Metrology Research Programme(EMRP) PartEmission project to compare the reference AEsdirectly to each other because accurate determination ofcharge concentration is a prerequisite for the realization ofa particle number concentration scale. This is the first AEcomparison (registered also as the EURAMET project 1244—comparison of AEs [18]) involving multiple NMIs and researchlaboratories with profound expertise in the field.

In the comparison reported here, participant laboratoriesbrought their reference AEs to the Aerosol PhysicsLaboratory of Tampere University of Technology (TUT)where measurements were conducted from 18 to 22 March2013. The aim was to investigate the validity of the unitrealization among laboratories and to demonstrate the bestmeasurement capability that can be achieved for particlecharge concentration. The instruments were operated by therepresentatives of the owner laboratories. The calibrationaerosol was generated using the single charged aerosolreference (SCAR [19]), which generates singly chargedparticles with known size. The measurement scheme andexperimental setup were carefully designed to minimize theuncertainty due to the comparison method. For examplespecial care was taken to ensure symmetric flow splittingof the aerosol such that the particle concentrations at theAE inlets were equal. The comparison measurementswere performed at different calibration particle sizes ranging

294

Metrologia 51 (2014) 293 R Hogstrom et al

Table 1. Participants in the study.

Country Organization Status Role

UK National Physical Laboratory (NPL) NMI CoordinatorFinland Centre for Metrology and Accreditation (MIKES)a NMI ParticipantFinland Tampere University of Technology (TUT)a Expert lab HostSwitzerland Federal Institute of Metrology (METAS) NMI ParticipantGermany Physikalisch-Technische Bundesanstalt (PTB) NMI ParticipantJapan National Institute of Advanced Industrial Science and Technology (AIST) NMI ParticipantEU Joint Research Centre (JRC) Expert lab ParticipantGermany Leibniz Institute fur Tropospharenforschung (TROPOS) Expert lab ParticipantUSA US Army Primary Standards Laboratory (APSL) Expert lab Participant

a MIKES and TUT participated with one instrument.

from 20 nm to 200 nm in order to reveal any particle sizedependent losses in the instruments. Measurements werealso performed at different charge concentrations from 0.16 ×10−15 C cm−3 to 2.72 × 10−15 C cm−3 using 100 nm singlycharged calibration particles (equal to particle concentrations1000 cm−3 to 17 000 cm−3). This range of particleconcentrations and sizes covers the scope of the calibrationrequirements set forth in the European regulation on exhaustemissions tests performed in the type approval of passengervehicles [20]. Additionally, measurements with sub-20 nmparticles were performed for studying the applicabilityof the experimental setup to comparisons with smallerparticles.

2. Reference standards

2.1. AE: operating principle and uncertainty sources

The AE is a flow-through instrument measuring chargeconcentration of aerosol particles. It comprises a FCUP, anelectrometer and flow control and measurement. The FCUPconsists of a high efficiency particulate filter enclosed in aconducting cup, which is surrounded by a grounded enclosureacting as a shield. The inner cup is connected to groundpotential only via a sensitive current-to-voltage amplifier, anelectrometer. As charged particles get collected in the filter,an equal but opposite electric charge (so-called image charge)passes through the electrometer so that the net charge onthe FCUP is zero. Therefore, the current measured by theelectrometer corresponds to the current carried by the particlesto the filter. Flow control and measurement is applied todetermine the gas volume rate corresponding to the electriccurrent. The aerosol particle charge concentration C iscalculated as

C = I

V · ηFCUP, (1)

where I is the electric current, V is the volume flow rate andηFCUP is the detection efficiency of the FCUP. The particlenumber concentration can be further calculated from thecharge concentration by dividing the charge concentrationwith the average number of elementary charges carried byparticles and the value of one elementary charge. Accordingto [12], particles generated by the SCAR generator are singlycharged with an uncertainty of only 0.16%. Therefore, the

number concentration is directly obtained in this comparisonby dividing the charge concentration with the elementarycharge.

It is fairly straightforward to obtain traceable electriccurrent and aerosol flow measurements through calibrations ofthe volume flow meters and electrometers. A more challengingtask is to determine the FCUP detection efficiency. Particlelosses may occur inside the FCUP through diffusion lossesin the inlet tube and/or deposition of particles outside thedetection enclosure. Also, particles might bypass the filterdue to leaks or penetrate the filter without being collected.Furthermore, leakage currents and/or improper grounding ofthe FCUP are possible sources of errors. The effect of theaforementioned defects on the FCUP detection efficiency ishard or even impossible to reliably and accurately quantify. Acomparison study is the only way of revealing such defects.However, with careful design these defects can be minimized.Including AEs of different design and manufacturer in thecomparison enables us to identify possible model dependentuncertainties in the results.

2.2. Participating laboratories and associated AEs

A total of nine laboratories participated in the comparison study(see details in table 1). MIKES and TUT participated with theAE of the SCAR instrument, which has been designed, builtand tested by TUT [19] and validated in co-operation withMIKES [12]. As a consequence, there were eight AEs to becompared as shown in table 2. All laboratories, except AIST,participated with their primary standards. The AIST standardis considered a secondary standard as it is calibrated against theJapanese primary standard [13]. All of these were commercialAEs or modified commercial AEs, except the MIKES–TUTstandard, which comprises a self-made FCUP and commercialcurrent and flow measuring instruments. Furthermore, theAIST secondary standard has traceability to the self-made AEof the Japanese primary standard. Each participant took care ofthe calibrations of the volume flow meter and electrometer usedin their AEs. Final results and associated uncertainties weresubmitted by the participants individually to the Coordinatorsome weeks after the comparison, to allow for further checksand recalibration at the participants’ laboratories.

295

Metrologia 51 (2014) 293 R Hogstrom et al

Table 2. Properties of participating laboratories AEs.

Electric currentParticipant AE model Status of standard measurement Flow measurement/control

NPL GRIMM 5.705 Primary Internal MKS 179/internalMIKES–TUT Self-made FCUP Primary Keithley 6430 Alicat Scientific MC-2SLPM-D/5 MMETAS TSI 3068B Primary Internal Vogtlin Red-y GSM-B4PA-BN00/internalPTB TSI 3068B Primary Internal InternalAIST TSI 3068B Secondary Internal InternalJRC Ioner EL-5030 Primary Internal Alicat Scientific MC-5SLPM-D/5 MTROPOS TSI 3068B Primary Internal InternalAPSL TSI 3068B Primary Internal Bios Defender 510-M/critical orifice

Figure 1. Experimental setup used in the comparison. The arrangement of AEs as seen from above is shown in the figure inside the dashedlines.

3. Comparison procedure

3.1. Experimental setup

The aerosol generator of the SCAR comprises fouroperational units which are used in numbered order forgenerating singly charged aerosol particles of known size.(1) Generation of NaCl seed particles by means of tube furnaceevaporation/condensation method. (2) Bipolar charging ofthe 10 nm to 12 nm seed particles. (3) Size classificationof the charged seed particles using a differential mobilityanalyser (DMA). (4) Condensational growth of seed particlesto desired size using di-octyl sebacate (DOS) as the workingfluid. The condensational growth preserves the particle charge(one elementary charge) acquired as result of steps 2 and 3 [12].A more detailed description of the operating principle of theSCAR is in [19].

The setup used in the AE comparison is depicted infigure 1. The output flow rate of the aerosol generator is bydesign constant at 2 l min−1. A dilution flow at the outputwas used for providing sufficient sample flow for the AEs

and to dilute the output particle concentration of the generator.Nitrogen was used as the dilution gas since it corresponds to thecarrier gas of the generated aerosol. Nitrogen was also used incalibrations of the flow meters of the participant AEs. Thiseliminates possible flow measuring inaccuracies caused bymismatch between calibration gas and the actual metered gas.

Special attention was paid to ensuring symmetric flowsplitting to the AEs. A static mixer was placed before aflow splitter (TSI flow splitter 3708) in order to guaranteespatially homogeneous particle concentration before the 4-wayflow splitter. In addition, a standard stainless steel Swagelok®

branch tee was used to further split the sample aerosol to theAEs (figure 1). The sample was drawn through the AEs at aflow rate of 1 l min−1 using their own pump or, if not available,using an external pump. All sample lines between the flowsplitter and the AEs were of equal length and diameter, therebythe diffusion losses in the sample lines were expected to besimilar. In this comparison study, the volume flow rate wasset and measured with reference to standard conditions (25 ◦Cand 1013 hPa). This simplifies the measurement setup as no

296

Metrologia 51 (2014) 293 R Hogstrom et al

Table 3. Summary of measurement points in the AE comparison. Measured combinations of charge concentration and particle size aremarked with x.

Nominal particle size/nmNominal charge Equivalent numberconcentration/10−15 C cm−3 concentration/cm−3 20 30 50 80 100 200

2.72 17 000 x1.60 10 000 x x x x xa x1.28 8 000 xa

0.64 4 000 x0.32 2 000 xa

0.16 1 000 xa

0 0 x

a Measured twice.

additional temperature and pressure sensors for each AE areneeded. The validity of the flow splitting was investigated forparticle sizes between 20 nm and 200 nm by means of separatemeasurements, which confirmed that the concentrations at theeight sampling ports agree within experimental uncertainties.However, the uncertainty of these measurements was added tothe uncertainty of the comparison results.

The temperature, pressure and humidity of the calibrationaerosol were also measured. These measurements were onlyperformed for monitoring purposes, as the flow rate throughthe AEs was specified as volume flow at standard conditions.In the comparison, the particle size distribution at the generatoroutput was measured before each calibration run using a SMPSconsisting of a TSI 3081 DMA, a TSI3080 controller and a TSI3776 CPC. The SMPS was operated at 1.5 l min−1 sample flowand 15 l min−1 sheath flow rate. Before each measurement,the nominal calibration concentration was adjusted based onthe reading of the TSI 3776 CPC. The uncertainty of the sizeassigned to the particles, which will be more significant atlower sizes, has not been estimated, as it is not critical to thecomparison.

3.2. Comparison measurement scheme

The comparison was performed with different particlesizes ranging from 20 nm to 200 nm in order to revealany particle size dependent losses within the instruments.These measurements were performed at a nominal chargeconcentration of 1.6 × 10−15 C cm−3 (equal to 10 000 cm−3

of singly charged particles) in order to achieve good signal-to-noise ratio and thus minimize the contribution of theuncertainty due to electrometer noise. Moreover, AEswere compared at different particle concentrations rangingfrom 0.16 × 10−15 C cm−3 to 2.72 × 10−15 C cm−3 (equalto 1000 cm−3 to 17 000 cm−3 singly charged particles) using100 nm calibration particles. This particle size was chosenbecause particle losses in the tubing are insignificant at thissize level. Measurements were repeated for the two lowestconcentrations where the AE noise is significant in order tofind out the repeatability that can be achieved with the appliedmeasurement cycle and also to compare the repeatabilityof AEs. The repeatability was also measured at a higherconcentration (1.28 × 10−15 C cm−3) where the AE noiseis less significant. Moreover, the zero concentration wasmeasured for detecting and compensating drifts in the AEs

zero current. A summary of the measurement points includedin this comparison is given in table 3.

The measurement cycle at each measurement pointconsisted of fifteen 1 min long periods of sampling at thenominal concentration (figure 2). Before and after eachof these periods, a 1 min period of zero concentration wassampled. This is necessary in order to compensate for thezero current (i.e. zero concentration) offset inherent for AEmeasurements. An estimate for the concentration duringa single measurement cycle was calculated by subtractingthe average value of two neighbouring zero concentrationperiods from the calibration concentration period (seeexpression inside brackets of equation (2)). After eachconcentration change, the instruments need some time tostabilize. Therefore, for each 1 min period, the concentrationwas calculated as an average of the last 30 s of data. As aresult, the charge concentration for one measurement pointwas calculated as an average of 15 zero corrected concentrationvalues as follows:

C =

15∑k=1

(Ck − C0,k+C0,k+1

2

)15

, (2)

where subscript k denotes the period number and C0 the zeroconcentration. The particles were switched on and off bytoggling on and off the voltage of the DMA inside of the SCAR.All AEs sampled at a sampling rate of 1 Hz.

3.3. Calculation of results

Each participant calculated their results according toequation (1). The uncertainty of the results was derived as(

u(C)

C

)2

=(

u(I)

I

)2

+

(−u

(V

)V

)2

+

(−u(ηFCUP)

ηFCUP

)2

+ u2(δCs), (3)

where u(δCs) is the standard uncertainty due to inhomogeneityin flow splitting which was added to the uncertainty reportedby each participant. Experimental investigations conductedbefore the comparison campaign showed that the particleconcentration at the sampling ports was equal within theexperimental standard uncertainty of 0.5%.

297

Metrologia 51 (2014) 293 R Hogstrom et al

Figure 2. Measurement cycle.

The comparison reference value Cref for each measure-ment point was determined as the weighted mean of the resultsreported by the participants (C1 . . . CN), in which the weightsare given as the inverse square of the associated standard un-certainties [21]:

Cref = C1/u2(C1) + · · · + CN/u2(CN)

1/u2(C1) + · · · + 1/u2(CN). (4)

The standard uncertainty of the reference value is given as

1

u2(Cref)= 1

u2(C1)+ · · · +

1

u2(CN). (5)

This method of determining the reference value is, however,only valid if the measurement results are consistent. Themeasurement results were checked for consistency using thechi-square test as described in [21]. The results were foundconsistent with 95% confidence for every measurement pointwhen the JRC results were excluded from the reference valuecalculations.

The results of the comparison are reported in terms of therelative difference di between the result Ci of laboratory i andthe comparison reference value Cref .

di = Ci − Cref

Cref× 100%. (6)

The corresponding standard uncertainty was calculated as

u(di) =√

u2(Ci) − u2(Cref)

Cref× 100%. (7)

The expanded uncertainty of di is calculated as

U(di) = 2u(di), (8)

which gives a 95% coverage assuming the results are normallydistributed [22].

The relative differences and their expanded uncertaintieswere used for identifying potential anomalies in the results as

|di | > U(di). (9)

Concentrations measured by the AE used by the JRC were10% to 30% lower than the other institutes’ concentrations.The deviation was size dependent such that it was larger for

larger particle sizes. A possible source of these deviationscould be the design of the FCUP inner structure, which mightcause a fraction of particles to get deposited onto the groundedinner walls of the FCUP and thus become undetected. Particledeposition mechanisms are known to depend on particle size[23], which would explain the observed size dependent losses.In fact, the FCUP has been redesigned in current models. Dueto this apparent shortcoming, it was decided not to report theJRC results, nor take them into account when calculating thecomparison reference value.

4. Results

4.1. Comparison results with different particle sizes

The results of the comparison measurements with differentparticle sizes are summarized in figures 3 and 4, and table 4.No clear dependence of the results on the particle size canbe identified in the results of figure 3. For all institutes, thevariations of the mean relative differences are smaller than1.1%. As a typical example of complete results obtained withdifferent particle size, figure 4 shows the results obtained withthe particle size of 100 nm and the charge concentration of1.6 × 10−15 C cm−3. It can be seen that both the AIST andthe PTB results show deviations from the reference value thatare close to the expanded uncertainty of the deviation. Similarresults were obtained with other particle sizes as shown intable 4. The uncertainties increase only slightly for smallerparticle sizes, due to uncertainties caused by diffusion lossesin the AE inlet. From table 4 it can also be seen that all themean results are within ±3% of the reference value and that allmeasurement results, except for the AIST result at 30 nm, agreewithin the estimated uncertainties. The AIST 30 nm result is,however, in good agreement with the rest of the AIST resultsand cannot therefore be considered an outlier.

4.2. Comparison results at different particle chargeconcentrations

The results obtained with different particle charge concentra-tions with the particle size of 100 nm are presented in figure 5and table 5. Similar to measurements at different particle sizes,the mean results agree within about ±3% at concentrationslarger than 1.28 × 10−15 C cm−3. For these measurements thedifferences between the institutes are very similar to the results

298

Metrologia 51 (2014) 293 R Hogstrom et al

Figure 3. Relative difference between the results obtained by participants and the reference value at different particle sizes with a nominalcharge concentration of 1.6 × 10−15 C cm−3 (equal to 10 000 cm−3).

Figure 4. Relative difference between the results obtained by participants and the reference value for a nominal particle size of 100 nm and anominal charge concentration of 1.6 × 10−15 C cm−3 (equal to 10 000 cm−3). Error bars show the expanded uncertainties (k = 2) of theresults.

Table 4. Comparison results at different particle sizes and a nominal particle charge concentration of 1.6 × 10−15 C cm−3 (equal to10 000 cm−3). The results are expressed as relative difference di from the comparison reference value together with associated uncertainties(k = 2).

Nominal particle size/nm

di /% 20 30 50 80 100a 200

NPL −0.1 ± 1.6 −0.5 ± 1.3 −0.3 ± 1.3 −0.3 ± 1.2 −0.3 ± 1.3 0.0 ± 1.5MIKES–TUT −0.4 ± 1.8 −0.6 ± 1.5 −0.2 ± 1.5 −0.4 ± 1.4 0.0 ± 1.5 −0.4 ± 1.5METAS 0.1 ± 1.5 0.1 ± 1.1 0.4 ± 1.2 0.1 ± 0.9 0.1 ± 1.1 −0.1 ± 1.1PTB −1.8 ± 3.1 −2.4 ± 3.0 −2.6 ± 2.9 −2.0 ± 3.0 −2.9 ± 2.9 −2.0 ± 2.7AIST 1.6 ± 2.1 1.8 ± 1.7 1.2 ± 1.7 1.2 ± 1.5 1.3 ± 1.6 1.1 ± 1.5TROPOS −1.5 ± 3.3 −1.5 ± 3.2 −1.5 ± 3.1 −1.3 ± 3.1 −1.3 ± 3.1 −1.0 ± 3.1APSL 2.1 ± 4.5 2.1 ± 3.2 1.8 ± 3.5 1.9 ± 3.3 1.5 ± 4.3 1.4 ± 2.9

a Results illustrated in figure 4.

shown in figure 4. Larger deviations and poorer repeatabilitycan be found in the results obtained with smaller concentra-tions. This is probably due to the increasing uncertainty ofcurrent measurements with smaller concentrations. Figure 6demonstrates this well: the pattern of the results for particlesize and concentrations of 100 nm and 0.16 × 10−15 C cm−3,respectively, is similar to that in figure 4, but the relative dif-ferences and associated uncertainties are larger. Similarly tothe measurements at different particle sizes, both the PTB andthe AIST results show deviations from the reference value thatare close, and at some measurement points larger than, the

expanded uncertainties. From figure 5 it can also be seen thatthe dependence on the concentration is very different betweenthe AEs at low concentration levels.

The relative differences of METAS and MIKES–TUTremain constant within 1% for the whole studied concentrationrange. Also, the NPL results were found independent ofparticle concentration, but the scattering of results was larger(2.2%) than for the METAS and MIKES–TUT AEs. Thisscattering is, however, within the estimated uncertainties.

Particle concentration-dependent trends can be identifiedin the results of APSL, AIST, TROPOS and PTB. From figure 5

299

Metrologia 51 (2014) 293 R Hogstrom et al

Figure 5. Relative difference between the results obtained by participants and the reference value for different particle charge concentrationsof 100 nm size particles.

Table 5. Comparison results with different nominal particle charge concentrations and a nominal particle size of 100 nm. The results areexpressed as relative deviation di from the comparison reference value and the associated uncertainties (k = 2).

Nominal particle charge concentration/×10−15 C cm−3

di /% 2.72 1.60 1.28 1.28 0.64 0.32 0.32 0.16a 0.16

NPL 1.3 ± 1.4 0.4 ± 1.3 0.3 ± 1.3 0.3 ± 1.3 0.5 ± 1.6 0.4 ± 1.3 −0.8 ± 1.3 −0.9 ± 2.0 −0.1 ± 1.1MIKES–TUT −0.4 ± 1.4 −0.4 ± 1.4 −0.4 ± 1.5 −0.2 ± 1.4 −0.3 ± 1.9 0.3 ± 2.7 −0.2 ± 3.1 0.0 ± 3.4 −0.6 ± 4.1METAS −0.6 ± 1.1 −0.4 ± 1.1 −0.4 ± 1.0 −0.4 ± 1.2 −0.4 ± 1.2 −0.6 ± 1.3 0.3 ± 1.5 −0.1 ± 1.8 −0.5 ± 2.6PTB −1.7 ± 2.0 −2.3 ± 3.1 −3.2 ± 3.7 −3.2 ± 3.8 −8.3 ± 7.2 −18.4 ± 14.1 −10.0 ± 14.5 −17.9 ± 23.2 −39.5 ± 30.6AIST 1.6 ± 1.5 1.5 ± 1.6 1.4 ± 1.5 0.8 ± 1.6 1.9 ± 2.9 3.4 ± 4.7 4.0 ± 3.8 3.5 ± 3.8 6.5 ± 6.5TROPOS −1.4 ± 2.1 −0.8 ± 3.3 −1.5 ± 4.0 −0.1 ± 4.1 −1.6 ± 8.2 −2.1 ± 14.1 −0.6 ± 14.5 −2.8 ± 23.2 −4.4 ± 30.6APSL 1.1 ± 3.1 1.3 ± 3.5 1.1 ± 3.6 1.1 ± 3.9 2.1 ± 7.2 −0.3 ± 12.4 2.6 ± 12.4 1.8 ± 19.3 5.8 ± 21.9

a Results illustrated in figure 6.

Figure 6. Relative difference between the results obtained by participants and the reference value for a nominal particle size of 100 nm and anominal charge concentration of 0.16 × 10−15 C cm−3 (equal to 1000 cm−3). Error bars show the expanded uncertainties (k = 2) of theresults.

it can be seen that for the first three of these the differencefrom the reference value increases as the particle concentrationbecomes smaller and reaches a value of 5% to 6% for thelowest concentration of 0.16 × 10−15 C cm−3. The differencesare, however, less than the stated uncertainties, except for oneof the AIST low concentration results. Overall, the AISTresults deviated more than the associated expanded uncertaintyat 0.32 × 10−15 C cm−3 and 2.72 × 10−15 C cm−3. Theseresults were, however, in good agreement with the rest of theAIST results indicating that these results cannot be consideredas outliers. The PTB results show a different behaviour

from the other AEs. The deviation from the reference valueincreases notably with decreasing particle concentrations atconcentrations smaller than 1.28 × 10−15 C cm−3 (equal to8000 cm−3). The different behaviour of the PTB AE in termsof repeatability and concentration dependency indicates thatthe results are potentially anomalous. In spite of this, the PTBresults were included in the reference value because no clearreason for this odd behaviour was found. Moreover, the PTBresults cannot be classified as inconsistent according to theoutcome of the chi-square test. Due to the large uncertainties,the PTB results have very small effect on the reference value.

300

Metrologia 51 (2014) 293 R Hogstrom et al

Figure 7. Relative difference between the results obtained by the participants and the reference value for particle sizes 6 nm to 20 nm.

For example, at the lowest concentration point, the referencevalue would shift upwards by less than 0.1% if the PTB resultswere excluded from the reference value calculations.

4.3. Additional measurements with sub-20 nm particles

Additional measurements were performed to find out theapplicability of the experimental setup for comparisons withsub 20 nm particles. At these particle sizes the diffusion lossesin the sampling lines become significant. This has an effect onthe maximum particle concentration that can be achieved andalso on the symmetry of flow splitting. Even small differencesin the dimensions, flow geometry and flow rates of the brancheswill have a relatively larger effect for small particles.

This is well demonstrated by the results shown infigure 7. The relative differences between the resultsof the institutes increase as the particle size gets smaller(institute identifications are not shown in the figure). Theobserved differences cannot exclusively be explained by theslightly increased uncertainties due to diffusion corrections.Because of diffusion losses in the sampling lines, the chargeconcentration at the sampling ports decreases from a nominalvalue of 1.6 × 10−15 C cm−3 (equal to 10 000 cm−3) for 20 nmparticles to 0.4 × 10−15 C cm−3 (equal to 2700 cm−3) for 6 nmparticles. Therefore, it is not clear to what extent the observedincrease of deviation is due size dependent losses inside theAEs or different AE responses to particle concentration or dueto uneven flow splitting between sampling ports.

5. Discussion

The relative differences between the participants are fairlyconstant over the particle sizes and concentrations studied inthis comparison. No indication of asymmetric flow splittingwas found suggesting that the observed deviations are probablydue to uncertainties of volume flow and electric currentmeasurements. If the particle concentration at the samplingports were asymmetric due to different particle losses inthe branches, then this asymmetry would depend on particle

size as the loss mechanisms are particle size dependent [23].Although the mean correction due to the asymmetry wasestimated to be zero for all AEs, the associated standarduncertainty of 0.5% was included in the uncertainty analysis.

With large particle concentrations, the most significantsources of uncertainty are related to the volume flow andflow splitting. The uncertainty related to volume flow isaround 1% for all institutes, depending on the method of flowcontrol and measurement, and calibration. The differences inthe uncertainties reported by the participants are to a largeextent caused by different calibration uncertainties of theelectric current measurement devices. At low particle chargeconcentration levels the uncertainty of current measurementbecomes the dominating uncertainty component. Increasingrelative calibration uncertainty and noise in electric currentmeasurements with smaller concentrations cause the observedpoor repeatability and scattering of results at low particleconcentrations.

Reasons for the observed non-linear behaviour of the PTBAE are not fully understood. PTB used the same type ofAE as METAS, TROPOS, AIST and APSL but showed verydifferent characteristics. Electric current calibration of thePTB AE performed at PTB before the comparison did not showsuch non-linearity and the same current calibration procedurewas performed at PTB for the TROPOS AE. Moreover,recalibration of the PTB AE after the comparison indicatedthat the AE response had not changed during the comparison.These points indicate that the non-linearity is not due to theelectric current calibration nor is it caused by transport of theAE. An error in the measurement of the zero current duringthe comparison might have caused the observed behaviour, asit would have a relatively larger effect on small current values(small concentrations). After cleaning and recalibration of theAE at the manufacturer, a more stable offset was observed,also interference seen as transient spikes in the current signalwas lower than during the comparison at TUT.

Although the results of the institutes show a fairly goodagreement, the PTB and the AIST results had a systematicdeviation from the reference value that was very close to,

301

Metrologia 51 (2014) 293 R Hogstrom et al

and at some measurements points larger than, the expandeduncertainty of the deviation. This does not, however, imply thatthe results are invalid. It only shows that a mutual agreement ofthe primary realizations of the charge concentration unit is notachieved at these measurement points. As the disagreementwas mostly observed at low particle charge concentrations,it might indicate that there are uncertainties related toelectric current measurements that have not been properlyaccounted for in the uncertainty evaluation. Electric currentmeasurements and calibrations in the femtoampere range arevery challenging and further improvements in these areas areneeded for accurately measuring non-linearity of electrometersand for better understanding the factors contributing to theuncertainty. Also, the contribution of noise to the uncertaintyat low particle concentrations can be further reduced byperforming repetitive measurements and/or applying extendedmeasurement cycles.

The anomalies observed in the comparison resultsemphasize the importance and need for inter-laboratorycomparisons. Even a careful validation of the measurementstandard (i.e. AE in this work) and calibration of instrumentsmeasuring the input quantities (in this case volume flowand electric current) do not guarantee valid realization ofthe derived measurement quantity, i.e. particle chargeconcentration. Besides particle charge concentration, alsoexact knowledge about average particle charge is needed forrealizing the particle number concentration unit. Next, asimilar comprehensive comparison involving CPCs is neededfor investigating the equivalence of complete particle numberconcentration realizations at different institutes taking alsointo account the particle generation methods. In fact, sucha comparison (registered as the EURAMET project 1252—Comparison of CPCs [24]) has already been initiated.

Overall, the charge concentration measurements of theparticipating institutes were found to agree within the stateduncertainties except for few measurement points, with theagreement being within about 3%. Additional measurementswith sub-20 nm particles show that comparisons in this sizerange require special considerations regarding sampling linesin addition to a careful characterization of asymmetry in flowsplitting in order to minimize the experimental uncertainty.Even so, agreement with 6 nm particles was still observedto be within about ±5%. This outcome is an importantstep towards internationally recognized traceability in particlenumber concentration measurements.

6. Conclusions

This paper reports the first comprehensive comparison ofaerosol particle charge concentration standards at NMI level.Eight laboratories from seven countries participated in thiscomparison with their aerosol electrometers. The comparisoncovered a particle size range of 20 nm to 200 nm andcharge concentrations from 0.16 × 10−15 C cm−3 to 2.72 ×10−15 C cm−3 (equal to 1000 cm−3 to 17 000 cm−3 singlycharged particles). The analysed results for all particlesizes and charge concentrations down to 1.28 × 10−15 C cm−3

(equal to 8000 cm−3) agreed within about ±3%. At lower

concentrations, and at sizes between 6 nm and 20 nm, theagreement was about ±5%. However, a few potentiallyanomalous results were observed indicating that furtherresearch is needed.

As a summary, the results reported here show thatthe charge concentration measurements performed by theparticipating laboratories were mostly equivalent within theirstated uncertainties (k = 2). This study indicates thatthe applied comparison protocol and measurement setup iswell suited for performing accurate comparisons of AEs ina way that anomalous results can be identified. Additionalmeasurements show, however, that short sampling lines andcareful characterization of asymmetry in flow splitting arenecessary for comparisons in the size range below 20 nm.

Acknowledgments

This work was supported by European Metrology ResearchProgramme (EMRP) jointly funded by the EMRP participatingcountries within EURAMET and the European Union, andalso by the Cluster for Energy and Environment (CLEEN Ltd,MMEA, WP 4.5.1).

References

[1] IPCC 2013 Fifth Assessment Report (AR5) (Geneva: TheIntergovernmental Panel on Climate Change)

[2] European Commission 2001 CAFE (Clean Air for Europe)programme

[3] NSF 2003 Workshop Report on ‘Emerging Issues inNanoparticle Aerosol Science and Technology (NAST)’(University of California Los Angeles, 27–28 June 2003)

[4] Valavanidis A, Fiotakis K and Vlachogianni T 2008 Airborneparticulate matter and human health: toxicologicalassessment and importance of size and composition ofparticles for oxidative damage and carcinogenicmechanisms J. Environ. Sci. Health C 26 339–62

[5] Kulmala M, Vehkamaki H, Petaja T, Dal Maso M, Lauri A,Kerminen V M, Birmili W and McMurry P H 2004Formation and growth rates of ultrafine atmosphericparticles: a review of observations J. Aerosol Sci. 35 143–76

[6] Agarwal J K and Sem G J 1980 Continuous flow, singleparticle counting condensation nucleus counter J. AerosolSci. 11 343–57

[7] Knollenberg R G and Luehr R 1976 Open cavity laser ‘active’scattering particle spectroscopy from 0.05 to 5 microns FineParticles—Aerosol Generation, Measurement, Samplingand Analysis ed B Y H Liu (London: Academic) pp 669–96

[8] Wang S C and Flagan R C 1990 Scanning electrical mobilityspectrometer Aerosol Sci. Technol. 13 230–40

[9] Keskinen J, Pietarinen K and Lehtimaki M 1992 Electrical lowpressure impactor J. Aerosol Sci. 23 353–60

[10] ISO/DIS 14644-12 Cleanrooms and Associated ControlledEnvironments: Part 12. Classification of Air Cleanliness byNanoscale Particle Concentration Cleanliness (Geneva:International Organization for Standardization) (still underdevelopment)

[11] European Commission 2008 Regulation No. 692Implementing and amending regulation (EC) No 715/2007of the European parliament and of the council on thetype-approval of motor vehicles with respect to emissionsfrom light passenger and commercial vehicles (Euro 5 andEuro 6) and on access to vehicle repair and maintenanceinformation Offic. J. Eur. Union L:199/1

302

Metrologia 51 (2014) 293 R Hogstrom et al

[12] Hogstrom R, Yli-Ojanpera J, Rostedt A, Iisakka I, Makela JM, Heinonen M and Keskinen J 2011 Validating the singlecharged aerosol reference (SCAR) as a traceable particlenumber concentration standard for 10 nm to 500 nm aerosolparticles Metrologia 48 426–36

[13] Sakurai H and Ehara K 2011 Evaluation of uncertainties infemtoampere current measurement for the numberconcentration standard of aerosol nanoparticles Meas. Sci.Technol. 22 024009

[14] Fletcher R A, Mulholland G W, Winchester M R, King R Land Klinedinst D B 2009 Calibration of a condensationparticle counter using a NIST traceable method Aerosol Sci.Technol. 43 425–41

[15] ISO/DIS 27891 Aerosol Particle NumberConcentration—Calibration of Condensation ParticleCounters (Geneva: International Organization forStandardization) (still under development)

[16] Schlatter J 2009 Final Report of EURAMET Project 1027(www.euramet.org)

[17] Yli-Ojanpera J, Sakurai H, Iida K, Makela J M, Ehara K andKeskinen J 2012 Comparison of three particle numberconcentration calibration standards through calibration of asingle CPC in a wide particle size range Aerosol Sci.Technol. 46 1163–73

[18] EURAMET Project 1244—Comparison of AerosolElectrometers (www.euramet.org)

[19] Yli-Ojanpera J, Makela J M, Marjamaki M, Rostedt A andKeskinen J 2010 Towards traceable particle numberconcentration standard: single charged aerosol reference(SCAR) J. Aerosol Sci. 41 719–28

[20] European Commission 2012 Regulation No 83 of theEconomic Commission for Europe of the United Nations(UN/ECE)—Uniform provisions concerning the approvalof vehicles with regard to the emission of pollutantsaccording to engine fuel requirements Offic. J. Eur. UnionL:42/1

[21] Cox M G 2002 The evaluation of key comparison dataMetrologia 39 589–95

[22] BIPM, IEC, IFCC, ILAC, ISO, IUPAC, IUPAP andOIML 2008 Evaluation of MeasurementData—Guide to the Expression of Uncertainty inMeasurement (Joint Committee for Guides inMetrology)

[23] Hinds W C 1999 Aerosol Technology: Properties, Behavior,and Measurement of Airborne Particles (New York: Wiley)pp 191–6

[24] EURAMET Project 1252—Comparison of CondensationParticle Counters (www.euramet.org)

303